Defect detection method and device, hyperparameter optimization method and device

Abstract

The defect detection method and device, and the hyperparameter optimization method and device. The hyperparameter optimization method first acquires a sample image set, then for each hyperparameter combination in multiple hyperparameter combinations, trains a target task model using the sample image set with the hyperparameter combination, and evaluates and records the performance index of the trained target task model when performing the target task. Finally, the hyperparameter combination corresponding to the trained target task model with the best performance index is selected as the final hyperparameter combination used to train the target task model. In the process of training the model, if the prediction error of the target task model for any sample image is less than the error threshold, the sample image is discarded and no longer participates in the subsequent round of training. This not only reduces the sample amount used in the hyperparameter optimization process and reduces the time consumption, but also enables the model to use the most important and representative samples to efficiently and accurately obtain the best hyperparameter combination.

Images